Audit Committee, Board of Director Characteristics, And Earnings Management

download Audit Committee, Board of Director Characteristics, And Earnings Management

of 26

Transcript of Audit Committee, Board of Director Characteristics, And Earnings Management

  • Journal of Accounting and Economics 33 (2002) 375400

    Audit committee, board of directorcharacteristics, and earnings management$

    April Klein*

    Stern School of Business, New York University, New York, NY 10012-1118, USA

    Received 20 October 2000; received in revised form 2 February 2002

    Abstract

    This study examines whether audit committee and board characteristics are related to

    earnings management by the rm. A negative relation is found between audit committee

    independence and abnormal accruals. A negative relation is also found between board

    independence and abnormal accruals. Reductions in board or audit committee independence

    are accompanied by large increases in abnormal accruals. The most pronounced effects occur

    when either the board or the audit committee is comprised of a minority of outside directors.

    These results suggest that boards structured to be more independent of the CEO are more

    effective in monitoring the corporate nancial accounting process. r 2002 Elsevier Science

    B.V. All rights reserved.

    JEL classification: K0; G3; M4

    Keywords: Earnings management; Corporate governance; Audit committee; Board of directors

    1. Introduction

    In December 1999, the NYSE and NASDAQ modied their requirements foraudit committees. Under the new standards, rms must maintain audit committeeswith at least three directors, all of whom have no relationship to the company that

    $I would like to acknowledge the helpful comments of S.P. Kothari (the editor), an anonymous referee,

    Eli Bartov, James Doona, Lee-Seok Hwang, Jayanthi Krishnan, Carol Marquardt and the participants at

    the Temple University and NYU workshops. The Ross Institute of the Stern School of Business provided

    nancial support.

    *Corresponding author. Tel.: +1-212-998-0014; fax: +1-212-995-4004.

    E-mail address: [email protected] (A. Klein).

    0165-4101/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.

    PII: S 0 1 6 5 - 4 1 0 1 ( 0 2 ) 0 0 0 5 9 - 9

  • may interfere with the exercise of their independence from management and thecompany (NYSE Listing Guide, Section 303.01(B)(2)(a)). These new requirementsrespond to the SECs call for improving the effectiveness of corporate auditcommittees in overseeing the nancial reporting process. One specic area of concernto the SEC is inappropriate earnings management by the rm dened as thepractice of distorting the true nancial performance of the company.1 The commonthread running through the SEC and stock exchange proposals is an implicit positiveconnection between earnings management and non-independent audit committees.Yet no study to date explicitly tests this assertion. The purpose of this paper is toundertake such a study.Using a sample of 692 publicly traded U.S. rm-years, I examine whether the

    magnitude of abnormal accruals (the proxy for earnings management) is related toaudit committee independence. After controlling for other determinants of abnormalaccruals and audit committee composition, I nd the magnitude of abnormalaccruals to be more pronounced for rms with audit committees comprised of lessthan a majority of independent directors. I also nd a negative association betweenabnormal accruals and the percent of outside directors on the audit committee.However, and contrary to the SECs intent, no difference in abnormal accruals isfound between rms with and without wholly independent committees.Given that the audit committees effectiveness is embedded within the larger

    corporate governess process, I also investigate whether abnormal accruals are relatedto other board characteristics. I nd signicantly negative associations betweenabnormal accruals and the percent of outside directors on the board, and forwhether the board is comprised of less than a majority of outside directors. Theseresults are harmonious to the audit committee ndings given that the auditcommittee reports to the board and that its members come from the full board.I also examine whether changes in board or audit committee independence are

    accompanied by changes in the level of abnormal accruals. The results dovetail withthe cross-sectional ndings. Firms that change their boards and/or audit committeesfrom majority-independent to minority-independent have signicantly largerincreases in abnormal accruals vis-"a-vis their counterparts. These ndings supportthe hypothesis that earnings management is negatively related to independent boardsand audit committees, but can also be a reection of a period of increasinguncertainty.The uniqueness of this paper versus other papers relating board characteristics to

    earnings management is that while previous papers either examine rms committingegregious nancial fraud (e.g., Dechow et al., 1996 and Beasley, 1996) or rms withincentives to overstate earnings (e.g., DeFond and Jiambalvo, 1994; Teoh et al.,1998a, b; Parker, 2000), I conduct my analyses on a sample of large, publicly tradedU.S. rms which a priori have no systematic upwards or downwards earnings

    1See SEC Chairman Arthur Levitts Address to NYU Center for Law and Business on September 28,

    1998, the SECs proposed rule 32-41987 published on October 8, 1999, and the nal rule on audit

    committee disclosure dated January 10, 2000 for use and denition of earnings management by the SEC.

    All three can be found on www.sec.gov.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400376

  • management. Thus, my results lend support to the exchanges and SECs assertionsthat for all large U.S. traded companies, independent audit committees and boardsare better able to monitor the earnings process.This paper also contributes to the growing literature on measuring abnormal

    accruals. Kasznik (1999), Bartov et al. (2000) and Kothari et al. (2001) demonstratethe importance of controlling for the rms earnings process when measuringabnormal accruals. Specically, they show that not controlling for reversals of prioryears accruals or growth patterns in earnings results in measurement error inabnormal accruals, which can lead to erroneous inferences. This problem isexacerbated if the measurement error is correlated with the partitioning variable(e.g., audit committee or board independence). The methods used throughout thispaper address these issues and suggest the necessity of using a matched-portfolio (orrm) technique as advocated by these authors.Section 2 discusses the stock exchange rules for audit committee composition.

    Section 3 develops the hypotheses about the expected associations between corporategovernance mechanisms and earnings management. Section 4 details the sampleselection criteria and contains descriptive statistics of the data. Section 5 discussesthe methodologies and econometric issues related to creating the adjusted abnormalaccruals. Section 6 contains cross-sectional analyses. Section 7 has the empiricalresults surrounding the associations between changes in board or audit committeecomposition and changes in adjusted abnormal accruals. The results in section 8support the inferences made throughout the paper. Section 9 concludes.

    2. NYSE and NASDAQ rules for audit committees

    Prior to December 1999, the stock exchanges and NASDAQ rules for auditcommittee composition were vague at best. Large, U.S. listed companies wererequired or encouraged to maintain audit committees with a majority or all membersbeing independent of management. However, no denition of independence wasgiven.In December 1999 the NYSE and NASDAQ modied their requirements by

    mandating listed companies to maintain audit committees with at least threedirectors, all of whom have no relationship to the company that may interfere withthe exercise of their independence from management and the company.2

    Simultaneously, the SEC adopted new rules to improve disclosures related to thefunctioning of corporate audit committees.3 Excluded from the audit committee are

    2See NYSE Listing Guide, Section 303.01(B)(2)(a); NASDAQ Market Listing Requirements Section

    4310(c)(26)(B). See also SEC Release Numbers 34-42231, 34-42232 and 34-42233, Adopting Changes to

    Listing Requirements for the NASD, AMEX, and NYSE Regarding Audit Committees. For the NYSE,

    foreign companies are excluded if their audit committee structure abides by the countrys rules. For the

    NASDAQ, companies with revenues less than $25 million are excluded. To be listed on the NYSE, rms

    must have at least $100 million of revenues.3See Release Number 34-42266, Adopting Rules Regarding Disclosure by Audit Committees,

    Including Discussions with Auditors Regarding Financial Statements.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 377

  • directors who are current employees, former employees within the last 3 years, havecross compensation committee links, or are immediate family members of anexecutive ofcer. In addition, the NASDAQ excludes any director who accepts non-director compensation from the rm in excess of $60,000 or whose employer receivesat least $200,000 in any of the past 3 years.These rules, however, are not steadfast. NASDAQ Rule 4310(c)(26)(B)(ii) allows

    the board under limited circumstances to appoint any non-current employee orfamily member to the audit committee. NYSE Section 303.01(B)(3)(b) gives theboard broader discretion in appointing directors with business relationships to therm. If the board determines that the independence of the director is notcompromised by the business relationship, then that director may serve on theboards audit committee. Thus, rms may maintain audit committees that are not100% independent.

    3. Corporate governance mechanisms and monitoring the rms nancial reportingprocess

    3.1. The role of board audit committees in resolving conflicts between management

    and outside auditors

    The audit committee primary oversees the rms nancial reporting process. Itmeets regularly with the rms outside auditors and internal nancial managers toreview the corporations nancial statements, audit process, and internal accountingcontrols.Although much emphasis has been put on the audit committees role in preventing

    fraudulent accounting statements (i.e., malfeasance of management or the outsideauditor), Magee and Tseng (1990), Dye (1991), and Antle and Nalebuff (1991) arguethat legitimate differences of opinion may exist between management and outsideauditors in how to best apply GAAP. Antle and Nalebuff (1991) conclude that thesedifferences result either in the auditor being dismissed or, more likely, in a negotiatednal nancial report. DeFond and Subramanyam (1998) postulate that clientlitigation risk may result in auditors preferring more conservative accounting choicesthan management for clients they perceive to be more risky. They present evidenceconsistent with this assertion for a sample of rms experiencing auditor changes.Nelson et al. (2000), using survey data, conrm that many reported earningsnumbers are negotiated. Overall, prior research suggests that the audit committeesrole as arbiter between the two parties is to weigh and broker divergent views of bothparties to produce ultimately a balanced, more accurate report. Equivalently, its roleis to reduce the magnitude of positive or negative abnormal accruals.The maintained hypothesis throughout this paper is that an independent audit

    committee is best able to serve as an active overseer of the nancial accountingprocess. I predict that audit committee independence will be negatively related toearnings management.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400378

  • 3.2. Board independence

    Several papers present evidence suggesting that effective governance and rmperformance increase with board independence (for example, see Brickley et al.,1994; Byrd and Hickman, 1992; Weisbach, 1988). Others document a negative linkbetween outside directors and the incidence of nancial fraud (see Dechow et al.,1996; Beasley, 1996). I test the assertion that a boards relative independence frommanagement is negatively related to earnings manipulation.

    3.3. Relationship investing

    Relationship investing encompasses all situations in which a large blockholdertakes an active, interventionist role in the rms economic processes. For large U.S.companies, relationship investing is often achieved by giving a large non-manage-ment shareholder or one of his representatives a seat on the board of directors. Beingon the boards audit committees gives these investors the opportunity to monitor therms nancial reporting process. I predict a negative relation between earningsmanagement and the incidence of at least one large (e.g., at least 5% shareholdings)outside director on the boards audit committee.

    3.4. CEO shareholdings

    Wareld et al. (1995) nd a negative relation between managerial stockholdingsand the absolute value of abnormal accruals. They interpret their results as beingconsistent with managerial shareholdings acting as a disciplining mechanism (Berleand Means, 1932; Jensen and Meckling, 1976). In a similar vein, Morck et al. (1988),and McConnell and Servaes (1990) nd a positive relation between Tobins Q andinside director shareholdings. However, Healy (1985) presents evidence that CEOsmanage earnings to maximize their bonuses. Aboody and Kasznik (2000) andYermack (1997) show that CEOs manage investors earnings expectations downwardprior to scheduled stock option award to increase the value of their awards, andNagar et al. (2000) present evidence that a rms discretionary disclosure ofaccounting data is related to the form of the CEOs compensation. If the CEOmanages earnings to increase his overall compensation, then there will be a positiverelation between CEO shareholdings and earnings management. Thus, no a prioriprediction is made.

    4. Data description

    4.1. Sample selection

    Data about boards and board audit committees are hand-collected from SEC-ledproxy statements. The initial sample contains all rm-years listed on the S&P 500 asof March 31, 1992 and 1993 with annual shareholder meetings between July 1, 1991

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 379

  • and June 30, 1993. Table 1 summarizes how the nal sample is constructed. Ieliminate 28 rm-years for rms domiciled outside the U.S. I also exclude 53 banks(SIC codes: 6000 to 6199) and 36 insurance companies (SIC codes: 63006411)because it is difcult to dene accruals and abnormal accruals for nancial servicesrms. Thus, all inferences in the paper are limited by the particular time period andsample selection.Schedule 14A (the proxy statement) requires rms to disclose each directors

    name, business experience during the last 5 years, other current directorships, familyrelationships between any director, nominee or executive ofcer, signicant currentor proposed transactions with management, signicant business relations with therm and number of shares held.4 Schedule 14A (Item 7(e)(1)) requires rms to statewhether they have a standing audit committee. If such a committee exists, rms arerequired to disclose its functions and responsibilities, its members, and the number oftimes the committee met during the last scal year. Four rm-years are eliminateddue to missing information about their audit committees.Compustat provides the earnings, cash ows from operations, and other nancial

    data needed to construct the abnormal accruals. CRSP and Compustat provide datafor many of the independent variables. One hundred and eighty rm-years areeliminated due to insufcient Compustat or CRSP data. The deletions ariseprimarily from two sources. First, I use a cross-sectional Jones regression model toestimate the unadjusted abnormal accruals for each sample rm. The modelsparameters are estimated by industry and I require each rm-year to have at leasteight observations with the same two-digit SIC code. Second, I use variants ofKaszniks (1999) matched-portfolio technique to adjust the rms abnormal accrualfor effects that are correlated with board and/or audit committee independence. Onetechnique ranks all Compustat rms into percentiles by the 10-year standard

    Table 1

    Sample used in analyses

    Firm-years

    Initial S&P 500 Sample for 19921993 1000

    Non-US rms (28)

    Banking rms (four-digit SIC code: 60006199) (53)

    Insurance rms (four-digit SIC code: 63006411) (36)

    Missing data on audit committees (4)

    Missing Compustat or CRSP data (180)

    Outlier for absolute value of abnormal accrual (7)

    Final sample 692

    4 Item 404(a) of Regulation SK of the 1934 Securities and Exchange Act denes signicant business

    transactions as any transaction between rm and director (or his/her place of business) that exceeds

    $60,000. Item 404(b) delineates the transactions as payments in return for services or property, signicant

    indebtedness, outside legal counseling, investment banking, consulting fees and other joint ventures.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400380

  • deviation of past total accruals and requires each sample rm-year to have anappropriate matched-portfolio for the adjustment. Thus, each sample rm must have11 years of Compustat data, ending in 1991 or in 1992. Finally, I remove sevenabnormal accrual outliers; each being more than 5 standard deviations from themean. In total, these requirements yield 692 observations.

    4.2. Corporate governance characteristics

    Consistent with prior research (e.g., Weisbach, 1988; Byrd and Hickman, 1992;Brickley et al., 1994), I classify directors as insiders, outsiders, or afliated (gray)with the rm. Insiders are current employees of the company. Outsiders have no tiesto the rm beyond being a board member. Consistent with the NYSE and NASDAQlisting requirements, afliated directors are past employees, relatives of the CEO, orhave signicant transactions and/or business relationships with the rm as dened byItems 404(a) and (b) of Regulation SX, or are on interlocking boards as dened byItem 402(j)(3)(ii) of Regulation SX.Table 2 reports data on board and audit committee composition. On average,

    58.4% of board members and 79.6% of audit committee members are outsiders.While no rm has a completely independent board, 73.8% of the rms in the samplehave boards in which the majority of directors are independent of management.5 Incontrast, 43.4% of audit committees are comprised of outside directors only and86.7% have a majority of independent directors.

    5. Adjusted abnormal accruals

    Any test of earnings management is a joint test of (1) earnings management and(2) the expected accruals model used.6 Acceptance or rejection of the null hypothesisof no earnings management cannot be disentangled from the key methodologicalissue of how well the chosen expected accruals model separates total accruals into itsunexpected (abnormal) and expected components.7 Moreover, Dechow et al. (1995),Guay et al. (1996), Kasznik (1999), Bartov et al. (2000), and Kothari et al. (2001)show that any proxy for abnormal accruals yields biased metrics if measurementerror in the proxy is correlated with omitted variables. More importantly, if theomitted variable is associated with the independent variable of interest or is within anon-random sample, then well-specied tests must include an adjustment for theomitted variable. Kasznik (1999) and Kothari et al. (2001) control for the correlatedvariable by using a matched-rm or portfolio technique to adjust the abnormalaccruals. I employ Kaszniks (1999) matched-portfolio technique.

    5 In 1992 and today, the NYSE, AMEX, and the NASDAQ required domestic listed rms to have a

    minimum of two outside or independent directors on their boards.6Many papers use the terms discretionary and non-discretionary accruals for expected and abnormal

    accruals.7Bernard and Skinner (1996), Guay et al. (1996), Dechow and Skinner (2000), and Kothari et al. (2001)

    contain excellent discussions of this issue.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 381

  • I begin by estimating a cross-sectional variant of the Jones (1991) expectedaccruals model for all rms k in industry j for year t:8 The model is

    ACCRjk;t=TAjk;t1 aj;t1=TAjk;t1 bj;tDREVjk;t=TAjk;t1

    gj;tPPEjk;t=TAjk;t1 ejk;t; 1

    where ACCRjk;t are total accruals for rm k in industry j in year t [Compustat item#18-Compustat item #304], TAjk;t1 are total assets [Compustat item #6], DREVjk;t isthe change in net sales [Compustat item #12], and PPEjk;t is gross property, plant andequipment [Compustat item #7]. The changes in revenues and PPE are used tocontrol for expected (i.e., economic-based) components in total accruals.9

    I use all rms on Compustat having the same two-digit SIC code for the rm-year.Industries with less than eight observations are dropped from the sample. Thenumber of rms used in each industry model ranges from 8 to 315. In total, 114 two-digit industry regressions for the 2-year period are estimated.

    Table 2

    Descriptive corporate governance data

    Whole board Audit committee

    Percentage of directors who are

    Insidersa (%) 22.5 1.4

    Outsidersb 58.4 79.6

    Afliatesc 19.1 19.0

    Percentage of rms with

    100% outside directors (%) 0 43.4

    Majority of outside directors 73.8 86.7

    Sample is for 692 US rms-years with audit committees listed on the S&P 500 as of March 31, 1992 and

    1993 with annual shareholder meetings between July 1, 1991 and June 30, 1993.a Insiders are current employees of the company.bOutsiders have no afliation with the company beyond for being directors.cAfliates are former employees, relatives of the CEO, board interlocks, or have signicant transactions

    and/or business relationships with the rm as dened by Items 404(a) and (b) of Regulation SX.

    8Other papers using this model include DeFond and Jiambalvo, 1994; Subramanyam, 1996; DeFond

    and Subramanyam, 1998; Becker et al., 1998; Teoh et al., 1998a, b; Peasnall et al., 1998; Guidry et al.,

    1999; DuCharme et al., 2001.9Bartov et al. (2000) test the efcacy of the unadjusted cross-sectional Jones model vis-"a-vis other cross-

    sectional and time-series expected accruals models. They conclude that the cross-sectional original Jones

    model is the only model consistently able to detect earnings management for a sample of rms receiving

    audit qualications. Dechow et al. (1995) and Guay et al. (1996) contrast the time-series Jones and

    modied Jones time-series models with other time-series models and conclude that the Jones models

    perform the best in detecting abnormal accruals.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400382

  • Next, for each rm-year ij; t in the S&P 500 sample, I calculate the unadjustedabnormal accrual dened as

    AACij;t ACCRij;t=TAij;t1faj;t1=TAij;t1

    bj;tDREVij;t=TAij;t1 gj;tPPEij;t=TAij;t1g; 2

    where aj;t; bj;t; and gj;t are the tted coefcients from Eq. (1).Table 3 reports descriptive statistics for the entire sample. The average (median)

    abnormal accrual is 0.004 (0.003). Testing for whether the mean abnormal accrualis different from zero yields a t-statistic of 0.54 (p-value=0.59). Forty-eight percent

    Table 3

    Descriptive statistics on accruals, and abnormal accruals

    Variable Mean Std. dev. Median Minimum Maximum %Positive

    Abnormal accruals (AAC)

    Unadjusteda 0.004 0.189 0.003 1.79 1.89 48p-value 0.588 0.425

    Abs(AAC)

    Unadjusteda 0.077 0.173 0.035 0.00003 1.886 100

    p-value 0.001 0.001

    Adjusted for s(total accruals)b 0.014 0.177 0.013 0.281 1.838 100p-value 0.001 0.001

    Total accrualsc 0.061 0.047 0.057 0.405 0.139 10Abs(total accruals)c 0.067 0.050 0.059 0.0002 0.4053 100

    Non-discretionary accrualsd 0.064 0.190 0.065 2.17 1.80 11Net incomee 0.056 0.075 0.048 0.298 0.576 84Operating cash owsf 0.117 0.077 0.107 0.126 0.483 97Assetsg (in $millions) 8,960 21,352 3,145 179 174,429 100

    Sample is for 692 US rms-years with audit committees listed on the S&P 500 as of March 31, 1992 and

    1993 with annual shareholder meetings between July 1, 1991 and June 30, 1993.

    Abs is the absolute value.aUnadjusted abnormal accruals (AAC) are the accruals prediction error, e.g., the difference between

    total accruals and estimated expected accruals. See Eq. (2).bAdjusted abnormal accruals is the unadjusted Abs(AAC) minus the Median Abs(AAC) for a portfolio

    of rms matched by the same standard deviation of the rms past 10 years total accruals.cTotal accruals are the difference between net income before extraordinary items (Compustat item #18)

    and cash ows from operations (Compustat item #308), deated by lagged total assets (Compustat

    item #6).dNon-discretionary accruals are estimated for each rm-year as the expected value of accruals based on

    Eq. (1).eNet Income is net income before extraordinary items (Compustat item #18) deated by lagged total

    assets (Compustat item # 6).fOperating cash ows is from the cash ows statement (Compustat item #308) deated by lagged total

    assets (Compustat item # 6).gAssets are total assets (Compustat item # 6).

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 383

  • of the abnormal accruals are positive; a sign test yields a p-value of 0.43. Thus, noevidence of systematic upward or downward earnings management activity isdetected. This nding most likely is due to the S&P 500 being a relatively randomsample with respect to earnings management incentives (see Healy and Wahlen,1999; Nelson et al., 2000). Because of this quality, I use the unsigned (absolute valueof) abnormal accruals as a proxy for the unadjusted combined effect of income-increasing and income-decreasing earnings management. Other earnings manage-ment studies using this measure are Wareld et al., 1995; Becker et al., 1998; Bartovet al., 2000.Next, I use Kaszniks (1999) matched-portfolio method to adjust the absolute

    value of the AAC. The adjustment for each sample rm is the median absolute valueof the AAC for a portfolio of rms matched by a variable that is correlated withboth the absolute value of abnormal accruals and board and/or audit committeeindependence. As Table 4, Panels A and B illustrate the absolute value of abnormalaccruals, audit committee independence and board independence are correlated at

    Table 4

    Spearman correlations (p-value in parenthesis)

    Panel A: Spearman correlations of absolute values of total accruals (TA)a, abnormal Accruals (AAC)b,

    adjusted abnormal accruals (AAAC)c, with possible correlated variables

    Variable Abs(TA) Abs(AAC) AAAC

    s(TA)a,i 0.33 (0.01) 0.19 (0.01) 0.01 (0.73)Abs(earnings)d 0.08 (0.03) 0.13 (0.01) 0.03 (0.40)Abs(earningst1) 0.07 (0.08) 0.11 (0.01) 0.00 (0.93)Earnings 0.27 (0.01) 0.15 (0.01) 0.00 (0.99)Abs(Dearnings) 0.26 (0.01) 0.20 (0.01) 0.13 (0.01)Abs(DCFO)e 0.18 (0.01) 0.20 (0.01) 0.11 (0.11)Abs(DTA) 0.06 (0.11) 0.08 (0.03) 0.06 (0.10)Debtf 0.05 (0.18) 0.14 (0.01) 0.06 (0.11)Log(Assets) 0.01 (0.83) 0.22 (0.01) 0.15 (0.01)

    Panel B: Spearman correlations of percentages of outside directors on audit committee (%Audout) and boards

    (%Outsiders on board) with possible correlated variables

    Variable %Audoutg %Outsiders on boardh

    s(TA)i 0.09 (0.01) 0.18 (0.01)Abs(earnings) 0.09 (0.01) 0.17 (0.01)Abs(earningst1) 0.10 (0.01) 0.20 (0.01)Earnings 0.08 (0.04) 0.20 (0.01)Abs(Dearnings) 0.05 (0.17) 0.03 (0.40)Abs(DCFO) 0.05 (0.17) 0.07 (0.07)Abs(DTA) 0.01 (0.80) 0.02 (0.53)Debt 0.06 (0.13) 0.16 (0.01)

    Log(Assets) 0.07 (0.05) 0.13 (0.01)

    MV/BVj 0.14 (0.01) 0.19 (0.01)Negative incomek 0.07 (0.05) 0.01 (0.74)

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400384

  • the 0.01 level with the standard deviation of past total accruals, the absolute value ofcurrent earnings, and the absolute value of last periods earnings, and at the 0.01 or0.04 level for the level of current earnings. Each earnings variable controls for therms inherent accruals or earnings process and is consistent with Kothari et al.(2001) assertion that this periods AAC is associated with the rms earnings process.In particular, each variable allows for this periods earnings to take into accountreversals of prior year accruals or growth trends in earnings.10

    Table 4 (continued)

    Panel C: Correlations among earnings, cash flows, and total accruals

    Abs(earnings) Abs(earningt1) Earnings Abs(Dearnings) Abs(DCFO) Abs(DTA)

    s(TA) 0.37 (0.01) 0.38 (0.01) 0.26 (0.01) 0.24 (0.01) 0.23 (0.01) 0.13 (0.01)Abs(earnings) 0.73 (0.01) 0.82 (0.01) 0.08 (0.04) 0.21 (0.01) 0.09 (0.01)

    Abs(earningst1) 0.71 (0.01) 0.09 (0.02) 0.21 (0.01) 0.07 (0.06)

    Earnings 0.11 (0.01) 0.16 (0.01) 0.04 (0.24)Abs(Dearnings) 0.20 (0.01) 0.20 (0.01)Abs(DCFO) 0.23 (0.01)

    aTotal accruals are net income before extraordinary items minus cash ows from operations. Total

    accruals are deated by lagged total assets.bAAC are abnormal accruals measured as the difference between total accruals and expected accruals

    using the cross-sectional Jones model (Eq. (1)).cAAAC is the absolute value of adjusted abnormal accruals measured as Abs(AAC) minus the Median

    Abs(AAC) for a portfolio of rms match on the standard deviation of past total accruals.dEarnings and earningst1 are net income before extraordinary items deated by lagged total assets for

    the current and lagged year, respectively.eCFO is cash ows from operations deated by lagged total assets.fDebt is long-term debt deated by lagged total assets.g%Audout is the percent of outside directors on the audit committee.h%Outsiders on board is the percent of outside directors on the board.is(TA) is the standard deviation of total accruals for the 10 years prior to the current year.jMV/BV is the market value of common equity divided by the book value of total common equity.kNegative income is equal to one if the rm has two past years of negative income before extraordinary

    items and zero otherwise.

    Abs is the absolute value

    10The positive correlation between the standard deviation of past total accruals and the absolute value

    of current abnormal accruals signies that rms with extreme accruals inherent in their business are more

    likely to have high discretionary accruals. The negative correlation between this variable and board

    (audit committee) independence supports Hermalin and Weisbachs (1998) prediction that rms

    with past extreme accruals are less likely to have an independent board. The positive correlations

    between the absolute value of current abnormal accruals and the earnings numbers are congruent to

    Kothari et al. (2001); Dechow et al. (1995); and Kaszniks (1999) ndings that this periods AAC is re-

    lated to last period return on assets. The negative associations between the earnings numbers and board

    or audit committee independence are similar to results reported by Hermalin and Weisbach (1991) and

    Klein (2002).

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 385

  • Initially, I match by the standard deviation of total accruals. The adjustedabnormal accrual for sample rm-year ij;t is

    AAACij;t;p AbsAACij;t;p Median AbsAACt;p; 3

    where AbsAACij;t;p is the absolute value of the abnormal accrual for rm-year ij;t;Median AbsAACt;p is the median absolute value of the abnormal accruals for aportfolio of Compustat rms, and p is the percentile ranking of the Compustat rmsstandard deviation of total accruals.To get the percentile rankings, I compute the standard deviation of total accruals

    for the 10 years prior to 1991 or 1992 for all Compustat rms with non-missing dataand assign them to percentiles based on their ordered rank. Each Compustatpercentile for 1991 contains approximately 87 rms; for 1992 each percentile hasapproximately 90 rms. Similarly, I compute each sample rms standard deviationof total accruals for the 10 years prior to 1991 or prior to 1992. Each sample rm ismatched by the percentile, p; and the AAAC is computed using Eq. (3).As Table 4, Panel A shows, AAAC is insignicantly correlated with the standard

    deviation of past total accruals (r 0:01; p-value=0.73), the absolute value ofcurrent earnings (r 0:03; p-value=0.40), the absolute value of past earnings(r 0:00; p-value=0.93), and current earnings (r 0:00; p 0:99). Thus, much ofthe measurement error due to these factors is removed. Note too from Panel C, thecorrelations among these variables range from 0.37 to 0.82, suggesting that thevariables capture much of the same processes. Finally, as shown in Table 3, theAAAC has a lower mean (median) of 0.014 (0.013) than the unadjusted Abs(AAC)(mean=0.077; median=0.035).

    6. Cross-sectional analyses

    6.1. Defining audit committee and board independence

    The maintained hypothesis is that more independent audit committees and/orboards are associated with smaller AAACs. One issue is determining independence.This is not a trivial exercise as the following discussion illustrates.I use three denitions of independence. The rst is to interpret audit or board

    independence as the percentage of outside directors on the audit committee or on theboard. This is a common denition used in the academic literature (e.g., Beasley,1996). However, as Hermalin and Weisbach (1991) show, the relation betweeneconomic outcomes (i.e., Tobins Q for Hermalin and Weisbach) and boardindependence may not be linear.A second path is to follow the NASDAQ and NYSEs guidelines and consider an

    audit committee independent only if all members are outside directors. Since noboards are comprised solely of outside directors, this denition is not feasible for theentire board. Under this denition, audit committees can function independently ifand only if all members are free from managerial inuence.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400386

  • A third denition of board or audit committee independence is for a majority ofmembers to be independent of management. Dechow et al. (1996), for example,dene a board as being inside-dominated if at least 50% of board members are rmofcers. The rationale behind this metric is that majority rule dominates board andcommittee actions.The differences among denitions, particularly between the 100% and 51% rules,

    can inuence how rms structure their boards. The exchange rules suggest thateffective monitoring requires boards to maintain audit committees with independentdirectors only. To achieve this, rms need to recruit independent directors and mayhave to increase board size (Klein, 2002). Yet, Yermack (1996) argues and showsthat rms with smaller boards (i.e., under 10 directors) are better performers. Famaand Jensen (1983) and Klein (1998) articulate that rms benet greatly by havinginsiders on the board since top managers bring in expertise about the organization tothe boards top-level decision making apparatus. These papers suggest that it may becostly for companies to maintain 100% independent audit committees. Thus, usingthe 51% (majority) denition may be a desirable alternative to many rms.

    6.2. Univariate models

    The dependent variable is the AAAC. Examination of its distribution reveals thatits shape is approximately lognormal. Accordingly, I estimate the regressioncoefcients by maximum likelihood using a NewtonRaphson algorithm on alognormal-dependent variable. Since there are sign predictions for all of the variablesexcept %CEO shareholdings, one-tailed tests are reported except for that variable.Table 5 presents coefcients for the univariate models.11 As predicted, I obtain

    signicantly negative coefcients for both board denitions and for the 51% auditcommittee independence denition. In contrast, the coefcients on the 100% auditcommittee independence denition, the percent of outsiders on the audit committee,the incidence of a large blockholder on the audit committee and the percent ofcommon equity owned by the CEO are insignicantly different from zero.The most statistically signicant coefcients are for the 51% board and audit

    committee cutoff levels. To check the sensitivity of the ndings, I re-estimate modelsusing cutoffs of 40% and 60%. Only the coefcient on the 60% cutoff of outsidedirectors is signicant at the 0.10 level or better. Taken as a whole, these resultssuggest that a majority outside membership may be a critical threshold for deriving ameaningful relation between director independence and the absolute value of theadjusted abnormal accruals.

    6.3. Multivariate models

    This section uses multivariate models relating board characteristics to abnormalaccruals. I control for other factors that may be related to the absolute value of

    11Pearson and Spearman correlations present similar results to the regression models. The one exception

    is Aud51%, which has a weaker relation (po0:10) than those presented in Table 5.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 387

  • abnormal accruals or board/audit committee independence. As Bartov et al. (2000)show, failure to control for confounding factors may result in falsely rejecting thenull hypothesis of no abnormal accruals when in fact the null is true.Previous studies suggest that the absolute change in the previous years income

    before extraordinary items divided by total assets, and nancial leverage (total debtdivided by total assets) are positively associated with earnings management, andpolitical costs (log of beginning years assets) are negatively related to earningsmanagement (see Wareld et al., 1995; Dechow et al., 1995; DeFond and Jiambalvo,1994; Becker et al., 1998; Dechow et al., 1996; Bartov et al., 2000). As the last columnof Table 4, Panel A shows, AAAC is signicantly correlated at the 0.01 level with theabsolute value of the change in earnings and with the log of rm assets.Klein (2002) reports that audit committee and board independence are

    signicantly related to market-to-book ratios, past negative earnings (two or moreprevious consecutive years), and rm size (log of beginning years assets). FromTable 4, Panel B, we see that %Audout is signicantly correlated with these threevariables; %Outsiders on the board is signicantly correlated with rm assets and

    Table 5

    Univariate models of absolute values of adjusted abnormal accruals (AAACs)a on corporate governance

    explanatory variables

    Explanatory

    variable

    Bd51%b %Outc Aud100%d Aud51%e %Audoutf 5%Block. on

    aud. comm.g%CEO sharesh

    Intercept 0.14 0.01 0.08 0.17 0.15 0.09 0.08

    (48.85)n (0.09) (36.70)n (36.93)n (13.50)n (71.13)n (50.73)n

    Coefcient 0.07 0.14 0.01 0.09 0.06 0.11 0.31(w2-Values) (9.36)n (6.72)n (0.36) (10.05)n (2.21) (2.27) (1.27)

    Sample is for 692 US rms-years with audit committees listed on the S&P 500 as of March 31, 1992 and

    1993 with annual shareholder meetings between July 1, 1991 and June 30, 1993.nSignicant at the 0.01 level.aAAAC is the absolute value of adjusted abnormal accruals measured as Abs(AAC) minus the Median

    Abs(AAC) for a portfolio of rms match on the standard deviation of past total accruals. The dependent

    variable is modeled as a lognormal distribution. The parameters are estimated by maximum likelihood

    using a NewtonRaphson algorithm.bBd51% takes on the value of one if the rms board has at least a majority of outside directors, and

    zero otherwise.c%Out is the percentage of outsiders on the rms board.dAud100% takes on the value of one if the rms audit committee has outside directors only, and zero

    otherwise.eAud51% takes on the value of one if the rms audit committee has at least a majority of outside

    directors, and zero otherwise.f%Audout is the percentage of outsiders on the rms audit committee.g5%Block. on aud. comm. is an indicator if an outside 5% blockholder sits on the boards audit

    committee.h%CEO shares is the percentage of common equity owned by the CEO.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400388

  • the market-to-book ratio, but not to negative income. %Outsiders is alsosignicantly correlated with debt.Table 6 contains the multivariate results. AAACs are negatively related at the 0.01

    level to whether the board or its audit committee has a majority of independentdirectors. In addition, the coefcients on the percentages of outsiders on the board oraudit committee are signicantly negative at the 0.10 and 0.01 levels, respectively.Thus, cross-sectionally, board and audit committee compositions are related toabnormal accruals. However, contrary to the intentions of the new guidelinespromulgated by the exchanges, there appears to be no meaningful relation betweenabnormal accruals and having an audit committee comprised solely of independentdirectors.The associations between earnings management and having a 5% outside

    blockholder on the audit committee or CEO shareholdings are unclear since thecoefcients for the former variable are signicantly negative for three models onlyand the coefcients for the latter variable are signicantly positive for two

    Table 6

    Multivariate models of absolute values of adjusted abnormal accruals (AAACs)a on Board and Audit

    Committee Composition (parameter estimates and w2-Values)

    Predicted sign Model 1 Model 2 Model 3 Model 4 Model 5

    Intercept 0.12 0.03 0.08 0.16 0.14

    (2.89)n (0.17) (1.07) (4.70)nn (3.57)nn

    Bd51%b 0.08(12.59)nnn

    %Outc 0.10(3.13)

    Aud100%d 0.003n

    (0.03)

    Aud51%e 0.12(18.89)nnn

    %Audoutf 0.10(6.11)nnn

    5%Blockholder 0.09 0.14 0.12 0.11 0.11on audit comm.g (1.44) (3.68)n (2.84)n (2.56)n (2.31)

    %CEO sharesh ? 0.21 0.48 0.38 0.47 0.39

    (0.58) (2.84)n (1.87) (3.04)n (1.97)

    MV/BVi ? 0.01 0.01 0.01 0.01 0.01

    (4.08)nn (7.61)nnn (6.52)nnn (7.47)nnn (5.96)nn

    Abs(DNI)j + 0.65 0.54 0.59 0.71 0.66(6.07)nnn (4.20)nn (5.01)nn (7.29)nnn (6.18)nnn

    Neg. NIk ? 0.09 0.09 0.09 0.09 0.10

    (2.06) (2.37) (2.40) (2.38) (2.47)

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 389

  • specications only. Finally, the control factors, with the exceptions of negativeincome and the log of assets, are signicantly different from zero.

    6.4. Summary

    In summary, the results indicate that after holding other factors constant, rmswith boards and/or audit committees composed of less than a majority ofindependent directors are more likely to have larger AAACs than their counterparts.A negative relation exists between AAACs and the percent of independent directorson the board and/or audit committee. In contrast, there is no evidence of asystematic association between having an all-independent audit committee andabnormal accruals.

    Table 6 (continued)

    Predicted sign Model 1 Model 2 Model 3 Model 4 Model 5

    Debtl + 0.22 0.15 0.18 0.19 0.19

    (9.21)nnn (4.31)nn (6.25)nnn (7.43)nnn (6.67)nnn

    Log(Assets)m 0.01 0.01 0.01 0.01 0.01(1.09) (1.74) (1.40) (0.96) (1.11)

    Sample is for 692 US rms-years with audit committees listed on the S&P 500 as of March 31, 1992 and

    1993 with annual shareholder meetings between July 1, 1991 and June 30, 1993.nSignicant at the 0.10.nnSignicant at the 0.05.nnnSignicant at the 0.01.aAAAC is the absolute value of adjusted abnormal accruals measured as Abs(AAC) minus the Median

    Abs(AAC) for a portfolio of rms match on the standard deviation of past total accruals. The dependent

    variable is modeled as a lognormal distribution. The parameters are estimated by maximum likelihood

    using a NewtonRaphson algorithm.bBd51% takes on the value of one if the rms board has at least a majority of outside directors, and

    zero otherwise.c%Out is the percentage of outsiders on the rms board.dAud100% takes on the value of one if the rms audit committee has outside directors only, and zero

    otherwise.eAud51% takes on the value of one if the rms audit committee has at least a majority of outside

    directors, and zero otherwise.f%Audout is the percentage of outsiders on the rms audit committee.g5%Block. on audit comm. is an indicator if an outside 5% blockholder sits on the boards audit

    committee.h%CEO shares is the percentage of common equity owned by the CEO.iMV/BV is the market value of the total rm over book value of assets, measured at the beginning of the

    scal year.jAbs(DNI) is the absolute value of the change in net income between years t 1 and t:kNeg. NI. is an indicator if the rm had two or more consecutive years of negative income, ending on

    the scal year prior to the shareholders meeting.lDebt is long-term debt divided by last years assets.mLog(Assets) is the natural log of the book value of assets.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400390

  • 7. Changes in abnormal adjusted accruals and changes in board andaudit committee composition

    Regressions using cross-sectional data describe associations between abnormaladjusted accruals and board and audit committee composition. In this section, Iscrutinize more directly the link between board and audit committee compositionand earnings management by testing whether the level of abnormal adjusted accrualschanges when board or audit committee structure changes.I identify 339 rms having the required data for both 1992 and 1993. DAAAC is

    dened as the AAAC for 1993 minus the AAAC for 1992. A positive numberindicates an increase in the level of abnormal accruals. The change in board or auditcomposition is the percent of outside directors on the board (audit committee) in1993 minus the percent of outside directors on the board (audit committee) in 1992.Greater independence is associated with a positive number. The alternativehypothesis is that boards (audit committees) moving towards more (less)independence will be accompanied by a decrease (increase) in abnormal accruals.12

    Both between-sample tests and regression analyses are used. In the between-sample tests, I determine whether the change in the level of abnormal accruals isstatistically different for rms experiencing changes in their board (audit committee)structures as compared to rms not experiencing the changes. The t-test assumesnormal distributions and tests for differences between means. When appropriate,adjustments to the denominator are made to accommodate statistical differences invariances between samples. The z-test does not assume normality and tests fordifferences between medians. One-sided tests are performed.As one would expect, and as Klein (2002) shows, there is a signicant relation

    between changes in board independence and changes in audit committeeindependence. For the sample, the Spearman correlation coefcient between thetwo variables is 0.45, signicant at the 0.01 level. Sixty-six rms had an increase inaudit committee independence between 1992 and 1993. For these rms, 48 also hadan increase in board independence, 9 had a decrease in board independence, and 9had no change in board independence over the same 2 years. Seventy-eight rms hada decrease in audit committee independence between 1992 and 1993. For these rms,49 had a reduction in board independence, 15 had an increase in boardindependence, and 14 had no change in board independence over the same 2 years.Panel A of Table 7 presents the between-sample tests. First, I examine changes in

    AAACs around changes in board independence. The 116 rms that reduced itspercentage of outsiders on the board experienced a 0.051 mean increase in abnormaladjusted accruals, compared to a decrease of 0.001 for other rms. The 24 rmswhose boards moved from a majority of outsiders to less than a majority of outsiders

    12Weisbach (1988) nds that poorly performing rms are more likely to change their boards towards

    more independence. Since audit committee independence is related also to board independence (Klein,

    2002); Weisbachs (1988) ndings suggest that a positive relation between the change in adjusted abnormal

    returns and audit committee independence may arise due to economic events. This phenomenon will bias

    my results away from the hypothesis of a negative relation between adjusted abnormal returns and board

    (audit committee) independence.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 391

  • Table 7

    Changes in adjusted abnormal accruals (DAAAC) following changes in board and audit committee composition for all rms in sample with two years of data(N 339)

    Panel A: Mean changes in adjusted abnormal accruals (DAAAC)a

    Change in

    %Outside directors

    on board or audit

    committee

    No change or

    opposite change in

    %Outside directors

    on board or audit

    t-Statistic for

    difference

    between

    means

    z-Statistic for difference between

    medians

    committee Chg NoChg

    Mean

    DAAACN Mean

    DAAACN #>

    median

    #>

    median

    z-stat

    Board composition

    Increase in %Out.b 0.002 133 0.021 206 0.82 109 60 1.44n

    Decrease in %Out. 0.051 116 0.006 223 1.89nn 67 102 2.06nn

    To >51% Out. 0.025 19 0.016 320 1.86n 9 160 0.22To o51% Out. 0.161 24 0.002 315 1.79n 18 151 2.55nnn

    Audit committee composition

    Increase in Audoutc 0.015 66 0.020 273 1.60n 30 139 0.68Decrease in Audout 0.035 78 0.007 261 0.79 41 128 0.55To 100% Audout 0.000 23 0.014 316 0.57 9 160 1.06To o100% Audout 0.049 28 0.010 311 1.91nn 18 151 1.59nTo >51% Audout 0.083 6 0.015 333 1.10 4 165 0.83To o51% Audout 0.131 10 0.001 329 1.74n 2 167 1.92nn

    A.

    Klein

    /J

    ou

    rna

    lo

    fA

    ccou

    ntin

    ga

    nd

    Eco

    no

    mics

    33

    (2

    00

    2)

    37

    5

    40

    0392

  • Panel B: Regression of DAAAC on changes in board or audit committee composition

    Variable Predicted sign Coefcient

    (t-statistic)

    Coefcient

    (t-statistic)

    Intercept 0.01 0.01

    (1.21) (0.99)

    D%Out 0.13(0.97)

    D%Audout 0.16(2.15)b

    Adjusted R2 0.00 0.01

    F -value 0.95 4.61nn

    nSignicant at the 0.01 level.nnSignicant at the 0.05 level.nnnSignicant at the 0.10 level.aAAAC is the absolute value of adjusted abnormal accruals measured as abs(AAC) minus the median abs(AAC) for a portfolio of rms match on the

    standard deviation of past total accruals.b%Out is the percentage of outside directors on the boardc%Audout is the percentage of outside directors on the audit committee

    A.

    Klein

    /J

    ou

    rna

    lo

    fA

    ccou

    ntin

    ga

    nd

    Eco

    no

    mics

    33

    (2

    00

    2)

    37

    5

    40

    0393

  • had a 0.161 average increase in abnormal adjusted accruals, against a 0.002 increasefor other rms. Both partitions produce statistically signicant differences in meansand medians.Next, I examine changes in audit committee independence. Firms moving from a

    wholly independent audit committee to a lesser independent committee produce anaverage increase in abnormal adjusted accruals of 0.049, compared to an averageincrease of 0.010 for other rms. The 10 rms whose audit committees shifted from amajority of outsiders to a less than a majority of outsiders had a 0.131 averageincrease in abnormal accruals, against an average increase of 0.001 for other rms.Again, both classications yield statistically different means and medians.In Panel B, I regress the change in AAAC on changes in board or audit committee

    independence. The coefcients on each variable are negative as expected. However,only the change in audit committee independence yields a statistically signicant t-statistic.

    7.1. Summary

    In summary, the results indicate that rms with boards and/or audit committeesthat move from majority-independent to a minority-independent structuresexperience large increases in AAACs in the year of the change compared to theircounterparts.

    8. Additional tests

    This paper uses an abnormal adjusted residual from the cross-sectional Jonesmodel as its measure of abnormal accruals. As do other papers in the literature, thismeasure is interpreted as being a proxy for earnings management. Since the results ofmy study depend on this measure, it is important to take reasonable steps to ensurethe metric is measuring abnormal accruals and not other rm characteristicsincluded in the model. Adjusting the Jones model for extreme accruals inherent ineach rms business (e.g., the standard deviation of past total accruals) is onemechanism. In this section, I perform other tests to ensure further that the inferencesdrawn thus far are valid.

    8.1. Modified cross-sectional Jones model

    Dechow et al. (1995) propose a modied Jones model in which

    AACjk;t ACCRjk;t=TAjk;t1 faj;t1=TAjk;t1

    bj;tDREVjk;t DRECjk;t=TAjk;t1 gj;tPPEj;t=TAj;t1g: 4

    The modication is that in the expected accruals model, revenue changes areadjusted for DRECjk;t; the change in receivables for year t: Dechow et al. (1995)calculate aj;t; bj;t; and gj;t from their original model (Eq. (1) of this paper) and use

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400394

  • these estimates in Eq. (3). Bartov et al. (2000) and Kothari et al. (2001) follow thismethodology. Other papers, e.g., Kasznik (1999), re-estimate aj;t; bj;t; and gj;t bymodifying Eq. (1) to include the adjustment for receivables. I use both specications.Using Eq. (4) as my measure of abnormal accruals, I follow the same methodology

    described in Section 5 and calculate an AAAC for each rm-year. The univariate andmultivariate results with specication are almost identical to those presented inTables 57. Kothari et al. (2001) also nd statistically insignicant differencesbetween using the Jones and modied-Jones cross-sectional models.

    8.2. Adjustment for current or past abnormal earnings

    Kasznik (1999) nds that measurement error for the signed abnormal accrual ofthe cross-sectional Jones model is positively related to net earnings. Kothari et al.(2001) nd the same result for last years net earnings. As Table 4, Panel C shows, thecorrelation between this years and last years abnormal earnings is 0.73 and thecorrelation between this years signed and unsigned earnings is 0.82. Further, allthree variables are highly correlated with the conditioning variable used thus far, i.e.,the standard deviation of past total accruals. Thus, it can be argued each issurrogating for aspects of the rms earnings process.To control for the possibility that my results reect these omitted correlated

    variables, I adjust the unsigned abnormal accruals by the median value of theunsigned abnormal accrual for three matched portfolios of Compustat rms. Thematchings are based on (1) the absolute value of this years earnings, (2) the absolutevalue of last years earnings, and (3) this years signed earnings. Using the unsigned(absolute) values preserves the research design. However, it assumes that the degreeof earnings manipulation is unaffected by rm performance (net earnings), anassumption that Kasznik (1999) and Kothari et al. (2001) show to be untrue for rmsthat a priori have incentives to manage earnings. Thus, I match rms by both signedand unsigned earnings.The procedure to compute the AAAC is the same as described in section 5. The

    difference is that the Abs(AAC) is adjusted by the Median Abs(AAC) for differentlymatched portfolios. Both univariate and multivariate tests akin to Tables 5 and 6 areconducted with the newly calculated dependent variables. For the multivariateprocess, the same independent variables are included.Table 8 contains the parameter estimates and w2-values for the univariate and

    multivariate procedures. Panel A has the results on matching by the absolute valueof current earnings. Panel B contains the results on matching by the absolute valueof last years earnings. Panel C matches by the current signed earnings. Since5%Blockholder on audit committee and %CEO shares are included in eachmultivariate regression, I present the range of parameter estimates and w2-valuesover the ve regressions.As Table 8 illustrates, the parameter estimates and statistical signicant levels are

    qualitatively the same whether I match by the past standard deviation of totalaccruals or the level of earnings. For example, in Panel A, the coefcients onAud51% are 0.096 (p 0:01) and 0.120 (p 0:01) for the unsigned current

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 395

  • earnings-matched univariate and multivariate models vis-"a-vis 0.09 and 0.12(p |:01;0.01) for the standard deviation-matched models in Tables 5 and 6. InPanel B, the coefcients on Aud51% are 0.095 and 0.127, respectively. In PanelC, the coefcients on Aud51% are 0.101 and 0.124, respectively. Thus, the overallresults are robust to whether I match by the standard deviation of total accruals, theunsigned levels of current or last years earnings, or the signed level of currentearnings.

    8.3. Regression approach

    A second method for controlling for omitted correlated variables is to includethem as separate regressors (see Bartov et al., 2000). The regression is the unadjustedAbs(AAC) on the independent variables included in Table 6 alongside the standarddeviation of total accruals, the absolute value of this years earnings or the absolutevalue of last years earnings.The advantage to using this technique over the matched-portfolio approach is that

    mismatching on the conditioning variable introduces noise into the dependentvariable, which weakens the power of the independent variables. The disadvantageto the regression approach over the matched-portfolio approach is that the formerimposes a cross-sectionally linear relation, with a xed coefcient on theconditioning variable. In contrast, the matched-portfolio method does not imposerestrictions these restrictions.

    Table 8

    Parameter estimates and chi-square values for board independence and audit committee independence for

    univariate and multivariate models in which the AAACs are calculated by matching portfolios of current

    or past earnings

    Variable Univariate procedurea Multivariate procedureb

    Panel A: Matching by absolute value of current earnings

    Board51% 0.068 (9.71)n 0.084 (14.06)n

    %Out 0.121 (5.36)nn 0.084 (2.08)Aud100% 0.000 (0.02) 0.006 (0.51)Aud51% 0.096 (11.41)n 0.120 (19.27)n

    %Audout 0.080 (3.77)nn 0.110 (7.85)n

    5%Blockholder on audit comm.c 0.136 (3.43)nn (0.093, 0.144)d (1.71, 3.24)e

    %CEO sharesc 0.288 (1.14) (0.222, 0.501)d (0.66, 3.23)e

    Panel B: Matching by absolute value of last years earnings

    Board51% 0.067 (9.28)n 0.076 (11.54)n

    %Out 0.123(5.61)nn 0.094 (2.78)nnn

    Aud100% 0.002 (0.01) 0.008 (0.16)

    Aud51% 0.095 (11.17)n 0.127 (22.21)n

    %Audout 0.078 (3.61)nn 0.118 (9.21)n

    5%Blockholder on audit comm.c 0.124 (2.76)nn (0.102, 0.150)d (2.04, 4.34)e

    %CEO sharesc 0.392 (2.07) (0.280, 0.557)d (1.07, 3.67)e

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400396

  • I estimate three regressions on the abnormal unadjusted AACs (untabulated).Each regression includes one of the three correlated variables. I do not include twoor all three together because of the high degree of co-linearity among the variables.The regression results with these specications exhibit many similarities to thosereported in Tables 5 and 6. Specically, the coefcients and signicance levels on theaudit and board independence variables remain qualitatively the same. Onedifference is that the coefcient on whether a large blockholder sits on the auditcommittee is signicantly negative (p-values between 0.01 and 0.02) for theregression approach. A second difference is that the coefcient on rm assets is

    Table 8 (continued)

    Variable Univariate procedurea Multivariate procedureb

    Panel C: Matching by this years earnings

    Board51% 0.073 (11.43)n 0.080 (13.02)n

    %Out 0.100 (3.68)nn 0.088 (2.30)Aud100% 0.004 (0.03) 0.005 (0.07)Aud51% 0.101 (12.79)n 0.124 (20.93)n

    %Audout 0.088 (4.76)nn 0.112 (8.34)n

    5%Blockholder on audit comm.c 0.121 (2.76)nnn (0.095, 0.144)d (1.82, 4.04)e

    %CEO sharesc 0.411 (2.34) (0.267, 0.534)d (0.97, 3.83)e

    nSignicant at the 0.01 level.nnSignicant at the 0.05 level.nnnSignicant at the 0.10 level.aUnivariate procedure is the maximum likelihood estimates of the abnormal earnings-adjusted accruals

    on the independent variable.bMultivariate procedure is the maximum likelihood estimates of the abnormal earnings-adjusted

    accruals on the independent variable and the other variables in Table 6.c 5%Blockholder on audit committee and %CEO shares are used singularly in the univariate

    procedures. They are included alongside the board or audit committee independence variable in the

    multivariate procedures.dThis is the range of parameter estimates over the ve multivariate procedures. For example,

    5%Blockholder on audit committee is included as an independent variable in the multivariate regressions

    alongside Board51%, %Out, Aud100%, Aud51%, and %Audout, respectively. For the regressions on

    current-earnings adjusted AACs, the parameter estimates on 5%Blockholder on audit committee range

    between 0.093 and 0.144.eThis is the range of chi-square values over the ve multivariate procedures. In Panel A: for

    5%Blockholder on audit committee, three chi-Square values are signicant at the 0.10 level, one is

    signicant at the 0.05 level and one is not signicant at the 0.10 level. For %CEO shares, two chi-square

    values are signicant at the 0.10 level and three are not signicant at the 0.10 level. In Panel B: For

    5%Blockholder on audit committee, three chi-Square values are signicant at the 0.10 level, one is

    signicant at the 0.05 level and one is not signicant at the 0.10 level. For %CEO shares, one chi-square

    values is signicant at the 0.10 level, one is signicant at the 0.05 level and three are not signicant at the

    0.10 level. In Panel C: For 5%Blockholder on audit committee, two chi-square values are signicant at the

    0.10 level, two are signicant at the 0.05 level and one is not signicant at the 0.10 level. For %CEO

    shares, one chi-square values is signicant at the 0.10 level, two are signicant at the 0.05 level and two are

    not signicant at the 0.10 level.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 397

  • signicantly negative at the 0.01 level for the regression approach. Thus, as Kothariet al. (2001) show, disparate approaches can produce different inferences.Nevertheless, the main result that abnormal accruals are related to board and

    audit committee independence remains robust to this technique.

    8.4. CEO on board nominating committee or executive compensation committee

    Klein (1998, 2000) and Shivdasani and Yermack (1999) demonstrate a negativeassociation between board independence and whether the CEO sits on the boardsnominating or executive compensation committee. Thus, I test if earnings manage-ment is positively related to whether the CEO sits on these committees by includingtwo indicators into the regression analysis performed in Table 6 (untabulated). CEOon nominating committee equals one if the CEO sits on the nominating committee orif the board has no nominating committee, and zero otherwise. CEO oncompensation committee equals one if the CEO sits on the compensation or if theboard has no compensation committee. The coefcient on CEO on nominatingcommittee is insignicantly different from zero, suggesting no relation between thisvariable and earnings management. In contrast, the coefcient on CEO oncompensation committee is signicantly different from zero at conventional levels,suggesting a positive relation between earnings management and whether the CEOsits on this committee. This nding is consistent with Aboody and Kaszniks (2000)and Yermacks (1997) research showing that CEOs manage investors expectationson earnings downwards prior to the issuance of stock option awards.

    9. Summary and conclusions

    This study examines whether audit committee and board characteristics arerelated to earnings management by the rm. The motivation behind this study is theimplicit assertion by the SEC, the NYSE and the NASDAQ that earningsmanagement and poor corporate governance mechanisms are positively related.Cross-sectional negative associations are found between board or audit committee

    independence and abnormal accruals. Most signicantly, strong results are foundwhen either the board or the audit committee has less than a majority of independentdirectors. Contravene to the new regulations, no signicant cross-sectionalassociation is found between earnings management and the more stringentrequirement of 100% audit committee independence. In addition, I nd that rmschanging their board or audit committee from having a majority to a minority ofoutside directors experience large increases in AAACs relative to their counterparts.The implications of this study depend on ones view of how accurately the AAAC

    measures earnings management. A generous interpretation is that, consistent withthe SECs and the exchanges concerns, a negative relation exists between auditcommittee or board independence and earnings management for all large tradedU.S. rms. This interpretation suggests that the exchange rules are reasonable,although maintaining a wholly independent audit committee may not be necessary.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400398

  • A less liberal interpretation is that rms with large accruals inherent in their earningsstructure are less inclined to have independent boards or audit committees.This paper uses sensitivity analyses to examine the robustness of the AAAC with

    respect to it being a good measure of abnormal accruals. The results are robust tofour separate measures of AAAC and to an alternative regression approach. Still,future work needs to be done on nding better measures of abnormal accruals.

    References

    Aboody, D., Kasznik, R., 2000. CEO stock option awards and the timing of corporate voluntary

    disclosures. Journal of Accounting and Economics 29, 73100.

    Antle, R., Nalebuff, B., 1991. Conservatism and auditorclient negotiations. Journal of Accounting

    Research 29, 3154.

    Bartov, E., Gul, F.A., Tsui, J.S.L., 2000. Discretionary-accruals models and audit qualications. Journal

    of Accounting and Economics 30, 421452.

    Beasley, M.S., 1996. An empirical analysis of the relation between the board of director composition and

    nancial statement fraud. The Accounting Review 71, 443465.

    Becker, C.L., DeFond, M.L., Jiambalvo, J., Subramanyam, K.R., 1998. The effect of audit quality on

    earnings management. Contemporary Accounting Research 15, 124.

    Berle, A.A., Means, G.C., 1932. The Modern Corporation and Private Property. MacMillan, New York.

    Bernard, V.L., Skinner, D.J., 1996. What motivates managers choice of discretionary accruals? Journal of

    Accounting and Economics 22, 313326.

    Brickley, J.A., Coles, J.L., Terry, R.L., 1994. Outside directors and the adoption of poison pills. Journal of

    Financial Economics 35, 371390.

    Byrd, J.W., Hickman, K.A., 1992. Do outside directors monitor managers? Evidence from tender offer

    bids. Journal of Financial Economics 32, 195222.

    Dechow, P.M., Sloan, R.G., Sweeney, A.P., 1995. Detecting earnings management. The Accounting

    Review 70, 193226.

    Dechow, P.M., Sloan, R.G., Sweeney, A.P., 1996. Causes and consequences of earnings manipulation: an

    analysis of rms subject to enforcement actions by the SEC. Contemporary Accounting Research 13,

    136.

    Dechow, P.M., Skinner, D.J., 2000. Earnings management: reconciling the views of accounting academics,

    practitioners, and regulators. Accounting Horizons 14, 235250.

    DeFond, M.L., Jiambalvo, J., 1994. Debt covenant violation and manipulation of accruals. Journal of

    Accounting and Economics 17, 145176.

    DeFond, M.L., Subramanyam, K.R., 1998. Auditor changes and discretionary accruals. Journal of

    Accounting and Economics 25, 3667.

    DuCharme, L.L., Malatesta, P.H., Sefcik, S.E., 2000. Earnings management: IPO valuation and

    subsequent performance. Journal of Accounting Auditing and Finance 16, 369396.

    Dye, R.A., 1991. Informationally motivated auditor replacement. Journal of Accounting and Economics

    14, 347374.

    Fama, E.F., Jensen, M.C., 1983. Separation of ownership and control. Journal of Law and Economics 26,

    301325.

    Guay, W.R., Kothari, S.P., Watts, R.L., 1996. A market-based evaluation of discretionary accrual models.

    Journal of Accounting Research 34, 83105.

    Guidry, F., Leone, A.J., Rock, S., 1999. Earnings-based bonus plans and earnings management by

    business-unit managers. Journal of Accounting and Economics 26, 85108.

    Healy, P., 1985. The effect of bonus schemes on accounting decisions. Journal of Accounting and

    Economics 7, 261279.

    Healy, P.M., Wahlen, J.M., 1999. A review of the earnings management literature and its implications for

    standard setting. Accounting Horizons 13, 365383.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400 399

  • Hermalin, B.E., Weisbach, M.S., 1991. The effects of board composition and direct incentives on rm

    performance. Financial Management 20, 101112.

    Hermaln, B.E., Weisbach, M.S., 1998. Endogenously chosen boards of directors and their monitoring of

    the CEO. American Economic Review 88, 96118.

    Jensen, M.C., Meckling, W., 1976. Theory of the rm: managerial behavior, agency costs, and ownership

    structure. Journal of Financial Economics 3, 305360.

    Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting Research

    29, 193228.

    Kasznik, R., 1999. On the association between voluntary disclosure and earnings management. Journal of

    Accounting Research 37, 5781.

    Klein, A., 1998. Firm performance and board committee structure. The Journal of Law and Economics

    41, 275303.

    Klein, A., 2000. CEO power, board independence and CEO compensation: an empirical investigation.

    Working Paper, New York University.

    Klein, A., 2002. Economic determinants behind variations in audit committee independence. Working

    Paper, The Accounting Review 77, 435452.

    Kothari, S.P., Leone, A.J., Wasley, C.E., 2001. Performance matched discretionary accruals. Working

    Paper, Sloan School of Management, Massachusetts Institute of Technology.

    Magee, R.P., Tseng, M.-C., 1990. Audit pricing and independence. The Accounting Review 65, 315336.

    McConnell, J.J., Servaes, H., 1990. Additional evidence on equity ownership and corporate value. Journal

    of Financial Economics 27, 595612.

    Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation. Journal of

    Financial Economics 20, 293315.

    Nagar, V., Nanda, D., Wysocki, P.D., 2000. Compensation policy and discretionary disclosure. Working

    Paper, Business School, University of Michigan.

    Nelson, M.W., Elliott, J.A., Tarpley, R.L., 2000. Where do companies attempt earnings management, and

    when do auditors prevent it? Working paper, Cornell University and George Washington University.

    Parker, S., 2000. The association between audit committee characteristics and the conservatism of nancial

    reporting. Working Paper, Santa Clara University.

    Peasnell, K.V., Pope, P.F., Young, S., 1998. Outside directors, board effectiveness, and earnings

    management. Working Paper, Lancaster University.

    Shivdasani, A., Yermack, D., 1999. CEO involvement in the selection of new board members: an empirical

    analysis. Journal of Finance 54, 18291853.

    Subramanyam, K.R., 1996. The pricing of discretionary accruals. Journal of Accounting and Economics

    22, 249282.

    Teoh, S.H., Welch, I., Wong, T.J., 1998a. Earnings management and the underperformance of seasoned

    equity offerings. Journal of Financial Economics 50, 6399.

    Teoh, S.H., Welch, I., Wong, T.J., 1998b. Earnings management and the underperformance of initial

    public offerings. Journal of Finance 53, 19351974.

    Wareld, T.D., Wild, J.J., Wild, K.L., 1995. Managerial ownership, accounting choices, and

    informativeness of earnings. Journal of Accounting and Economics 20, 6192.

    Weisbach, M., 1988. Outside directors and CEO turnover. Journal of Fniancial Economics 20, 431460.

    Yermack, D., 1996. Higher market valuation of companies with a small board of directors. Journal of

    Financial Economics 40, 185212.

    Yermack, D., 1997. Good timing: CEO stock option awards and company new announcements. Journal

    of Finance 52, 449476.

    A. Klein / Journal of Accounting and Economics 33 (2002) 375400400

    Audit committee, board of director characteristics, and earnings managementIntroductionNYSE and NASDAQ rules for audit committeesCorporate governance mechanisms and monitoring the firms financial reporting processThe role of board audit committees in resolving conflicts between management and outside auditorsBoard independenceRelationship investingCEO shareholdings

    Data descriptionSample selectionCorporate governance characteristics

    Adjusted abnormal accrualsCross-sectional analysesDefining audit committee and board independenceUnivariate modelsMultivariate modelsSummary

    Changes in abnormal adjusted accruals and changes in board and audit committee compositionSummary

    Additional testsModified cross-sectional Jones modelAdjustment for current or past abnormal earningsRegression approachCEO on board nominating committee or executive compensation committee

    Summary and conclusionsReferences